Hi Atsushi, Thanks for your reply.
Let me just ask if there is a schedule for further development on this section? Regards, Vilius. On 10 May 2010 05:58, Atsushi Eno <atsushi...@veritas-vos-liberabit.com>wrote: > Hello, > > I will first stabilize existing stuff as well as fill missing > functionalities, and there is a lot to do. There isn't any kind of > optimization. I'd rather welcome your contributions and won't work on it so > far. > > Atsushi Eno > > > > On 2010/05/08 0:14, Vilius Adamkavicius wrote: > >> >> We have implemented a simple self hosted web service in .NET using >> ServiceHost. The Client is a simple Silverlight app that polls for updated >> info every 5 seconds. The return messages tend to be no more that 1-10K. >> >> We ported the web service (self-hosted) to MONO 2.6. This seemed to go >> perfectly and it stood up to stress testing in devel...then we deployed it. >> Now we find that the service is unstable and can just hangs after anywhere >> between 2 minutes and 2 days. >> >> IP traces show that when it hangs the XML Post gets (request) gets >> received by the webservice but nothing is ever transmitted back. Also when >> it hangs netstat shows a persistent active connection on the port to which >> it is allocated. >> >> In an attempt to diagnose this we looked at the port behaviour using >> netstat. We looked at this while the MONO implementation of the webservice >> was running without problem; we compare this to the active port connection >> behaviour of the windows/.NET implementation- what found was worrying. >> >> When we run a single client and connect the windows/.NET version of the >> webservice we see *one port connection that persists*. >> >> When we run a single client and connect the MONO version of the webservice >> we see it holds around *12-16 active port connections at any one time*...and >> these seem to be continuously dying and regenerating. >> >> We postulated that each call to the webservice was making a new connection >> (unlike Windows); each connection would be used once and then timeout. Given >> the rate that the client polls we estimated that the timeout must be around >> a minute...then we found this in the MONO DefaultCommunicationTimeouts.cs >> >> Vilius: private DefaultCommunicationTimeouts () >> { >> close = open = send = TimeSpan.FromMinutes (1); >> receive = TimeSpan.FromMinutes (10); >> } >> >> This strongly supports our theory... >> >> The question is why does it keep creating a new connection for each call? >> Can the MONO 2.6 implementation of ServiceHost maintain a single connection >> for an application session (like Windows .NET) ? What is the ramification of >> this? >> >> >> _______________________________________________ >> Mono-devel-list mailing list >> Mono-devel-list@lists.ximian.com >> http://lists.ximian.com/mailman/listinfo/mono-devel-list >> >> > > -- Regards, Vilius.
_______________________________________________ Mono-devel-list mailing list Mono-devel-list@lists.ximian.com http://lists.ximian.com/mailman/listinfo/mono-devel-list